That is a total guess figure from me. But I know when I throw an album at Picard I’ll get that kinda rate. I find it interesting when tagging a compilation album I have just ripped from a CD. If I use Lookup, it will usually hit the exact album. Or maybe it picks an alternate edition of the release and I need to adjust my choice to a different year\country.
If I hit Scan it can be fascinating to see a scatter of albums it finds to do the matches with. All them will be correct recordings, but sometimes some very unrelated albums.
Go find one or two albums in your collection that are Compilations. NOW 42, Kylie’s Greatest Hits, stuff like that. And use those to test with. That will show you the kinda results you are getting. I expect most will find a good original album or single, but a handful will jump onto an alternate compilation instead.
What you are doing is not inappropriate - just bending the original design spec. Its something pretty common I’ve seen done a lot. Not everyone wants “albums” in their music collections. They just want a huge playlist of music to dive into. We all have our own ways of enjoying music.
You want to be able to say “Play me tunes from the 1980s” and not care that your actual track was on a compilation CD from 2004. This is why I think there is a real call for an actual addon for Picard to help refine tagging of that style.
Ah… those people. No idea what all that is about, but then I ain’t a scrobbler. I’m one of those weirdos who plays whole albums at a time. In original track order. Ripped from CDs.
To be slightly more back on topic - what made me initially read this thread was that fascinating finding that Picard runs faster with a monitor attached. The way the GPU is being used for the AcoustID maths is interesting - “It runs faster with a monitor attached” almost sounds comical.
I understand that. It’s all about the maths. Good to see it getting used for more general stuff than just gaming. Running 3 times faster for you is pretty impressive. Funniest is that this is built into the libraries and the devs didn’t even have to request it - this just happens.
Computers are dumb maths crunchers that do as they are told. Real good at repetitive stuff or pattern matching like Picard is hitting. The time it takes Picard to hammer through your 4700+ files is a huge time saver for you.
Where computers are dumb and lack real knowledge is they do as they are told. So if they have bad data, they just feed you that bad data unknowingly. (See other conversations on “AI” and the idiocy that can be fed from ChatGPT)
This is why it can be worth doing a bit of a double check over what Picard creates so you don’t get some surprise music played to you one day
Only for certain, highly math-oriented workloads. There is a reason we still put conventional, general-purpose CPUs into our systems.
Audio decoding is so cheap that offloading it to the GPU would likely be slower. I don’t think the fingerprinting used by Picard makes any use of the GPU, the speedup likely comes from some secondary effect (e.g. unlocking higher CPU frequencies, power limits, or additional system RAM).
